基于运动矢量的鲁棒视频水印算法  

Investigation on video encoding and decoding and watermarking based on motion vector

作　　者：朱德庆[1,2] 金文光[1] 

机构地区：[1]浙江大学信息与电子工程系,浙江杭州310027 [2]杭州师范大学现代教育技术中心,浙江杭州310036

出　　处：《浙江大学学报（理学版）》2010年第3期286-290,共5页Journal of Zhejiang University（Science Edition）

摘　　要：针对AVS-M标准提出了一种水印嵌入方法,直接利用AVS-M编码器中解码的数据进行边缘检测确定边缘块与非边缘块,把边缘块作为水印添加候选块,有效控制了编码器的开销.选择P帧中具有最大运动矢量幅值的宏块嵌入水印,水印嵌入通过修改运动估计搜索得到的运动矢量值实现,因水印嵌入修改运动矢量会对后续编码块产生影响,给出了AVS-M编解码器严格的补偿方案,补偿后的水印信息均能正确提取,证明了这种方法的有效性.针对运动矢量水印检测要求严格同步的缺点,提出了利用最大相似序列向量检测水印的方法,对虚警、漏警或攻击造成的水印信息丢失具有良好的检测效果.According to AVS-M standard a new watermark embedding algorithm was proposed,the decoded data of encoder of AVS-M was used to detect edge and decide whether a macroblock is edge-block or not.Edge-macroblocks in P frame with the biggest motion vector magnitude were selected to embed watermark.The influence of modified motion vector on the following blocks was carefully analyzed and compensation methods were given.Various format video sequences were used to test the algorithm which showed the compensation was efficient.Watermark embedding methods based on motion vector require synchronous detection and can not resist delete or insert attacks,a new detecting way named Maximum-Likelihood sub-sequence vector detection is given to resolve the problem.

关 键 词：运动矢量 数字水印 补偿 相似度检测 AVS-M 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]